Download Forecasting and Nowcasting Major Earthquakes

Forecasting and Nowcasting Major Earthquakes
An Automated Cloud-Based Approach
John Rundle
Distinguished Professor, University of California
Chairman, Open Hazards Group
Topics:
Science
Applications
Tohoku, Japan
Earthquake and Tsunami
March 11, 2011
How do we estimate risk
from these events?
What does this mean for the
insurance industry?
Pricing Earthquake Risk: Current Practice
Science
Engineering
Earthquake
forecast
model
Ground
shaking
model
Poisson
Log Normal
BPT
Fault models
Pattern Analysis
AMR
ETAS
NTW
Nowcast
Etc.
Attenuation
PGA/PGV
Lquifaction
Surface geology
Basin structure
Etc.
Economics
Products
Economic
loss model
Pricing of
insurance &
financial
securities
EP curves
Building damage
Socioeconomic data
Supply chain disruption
Lost income
Lifelines
Etc.
CAT Bonds
SPVs
Warrants
Traders
Hedge Funds
Options
Muni Bonds
Equities
Etc.
The Science
Nowcasting and Forecasting
Nowcasting Earthquakes
JBR et al. (2016)
Nowcasting
Forecasting is a probability of future activity in the hazard (earthquake) cycle
Nowcasting describes the current state of the hazard cycle
The term “Nowcasting” was first used to describe the current state of the
economic/business cycle
Earthquake Cycle Example: Nankai Trench, Japan
M Ando, Tectonophysics, v27, p112 (1975)
• Data from historic writings in Japan
• The basic idea of the earthquake “cycle”
started in Japan using historical data
Elastic Rebound on a Fault
Report of the 1906 San Francisco Earthquake Investigation (1910)
Fault Stress 
Cycle of earthquakes was related to the cycle of stress release and recovery
Time, t
Fault Slip
T
t1 t2
Time, t
Harry Fielding Reid (1859-1944)
Problem: How do we measure the state of stress in time??
Elastic Rebound using Proxy Data
Example:
Accumulation of Small Earthquakes Between Large Earthquakes
“Large” EQ: MLarge > 6
“Small” EQ: 6 > MSmall > 3
Natural Time:
Cumulative Number
of Small Earthquakes
Since the Last Large
Earthquake
Time, t
Accumulating small earthquakes are readily observable data unlike stress
The number of small earthquakes between large earthquakes has
a stable average ratio (Gutenberg-RichterRelation)
The count of small earthquakes since the last large
earthquake can be viewed as a marker for the hazard level
Example: Nowcasting the Nankai Trench
Earthquake Potential Score, Nankai Trench
Computed 3/31/2017. Depths < 50 km
Example: Nowcasting the Sanriku Trench
Earthquake Potential Score, Sanriku Trench
Computed 3/31/2017. Depths < 50 km
Sanriku
Nowcasting Web App for Global Megacities (Under Development)
HTML-Javascript-CSS-Python
Earthquake Forecasting
Current Practice
• Expert elicitation is frequently used in forecasting, meaning that
backtesting is not possible
• Most/many current forecasts use time-independent Poisson statistics
in forecasting
• Poisson forecasts have the property that they have no memory of
past events
• An example is the current UCERF3 forecast for California, which has
not been backtested
From Nowcasts to Forecasts
Forecasting with the Natural Time Weibull Method
JR Holliday et al. (2014)
• We begin by counting small earthquakes since the last large earthquake (Nowcasting)
• We build on the Nowcast by projecting the count forward in time using the current rate
of small earthquake activity
• We combine these ideas with Weibull (1952) statistics, which are commonly used for
engineering failure analysis
• The result is a fully automated computation of probability of future large earthquake
occurrence
• Automation allows backtesting and optimization
• We have built this technology into a series of automated cloud-based web sites:
www.openhazards.com
Open Hazards PSHA Cloud Computational Framework
User Interface (UI)
Javascript, Python, HTML
XML/JSON, Web Services
Global
Data
Cloud-Based PSHA Compute Engine
Database
Middleware
(PHP, MySQL)
Earthquake
forecast
model
GMPE
model
Damage
model
Automated
forecasting
and risk
analysis
allows
rigorous
backtesting.
Expert
elicitation
generally does
not allow
rigorous
backtesting
Example: Vancouver Island Earthquakes, NTW Forecast
Latest Significant Event was M6.6 on 4/24 /2014
JR Holliday et al. (2014)
Chance of M>6 earthquake in circular region
of radius 200 km for next 1 year.
Data accessed 4/26/2014
m6.6
11/17/2009
Observed
Behavior
m6.0,6.1
9/3,4/2013
Idealized
Expected
Behavior
Time, t
Optimized 12 month California-Nevada forecast:
Probabilities (%) vs. Time for Magnitude ≥ 6
Forecast as of 2011/11/30
Automating
Forecasts
Allows
Backtesting
and
Optimizing!
Scatter Plot
1997-present
Observed Frequency vs.
Computed Probability
Optimal California-NV Forecast
Optimal forecasts via backtesting, using common
validation and verification testing procedures.
Bootstrap Error
Analysis
Temporal
Receiver Operating
Characteristic
1997-present
Applications/Products
The Open Hazards Group
Risk Analysis and Management Reimagined
Risk: Applications and Products
The Open Hazards Group
• Public web site
• NTW forecasts web app
• Ground shaking web app
• Structural damage factor web app
• Residential seismic safety reports
• Commercial seismic safety reports
• Cat bond analysis
• Natural hazard disclosures (required in California for transfer of
property)
• Financial trading models for hedge funds
Automated Forecasting
in the Cloud
www.openhazards.com
Automated California Forecast – Updated Nightly
Forecasting in a Circular Region
Region
Selection
Tool
Ground Shaking Tool
Shaking
Intensity
Tool
Residential Seismic Safety Report
Residential Seismic
Safety Report
Risk Calculations: EP Curves
Residential Seismic Safety Report
Automated
Commercial
Portfolio
Evaluator
Pages from Sample Commercial Report
Natural Hazards Disclosures
for Transfer of Residential Real Estate in California
Kibou Notes 2012
Japan Railways – GC Securities
EP Curve from Investor Presentation
Kibou Notes 2012
Annualized Forecast for
1 Year from 4-11-2012
Annualized Baseline
before 3-11-2011
EP Curve Computed by Open Hazards Group using NTW Forecast
Hedge Fund Strategies: Miller Value Partners
Thank you for your attention
The Open Hazards Group
www.openhazards.com
Distinguished Professor of Physics and Geology, University of California, Davis
Co-Founder of Open Hazards Group and Chair of the Board, Davis, California
Executive Director Emeritus of the APEC Cooperation for Earthquake
Simulations (ACES)
A Senior Advisor to the Association of Pacific Rim Universities (APRU)
Visiting Professor at Tohoku University at the APRU Multihazards Hub, Tohoku
University, Sendai, Japan
John was Chair (1994-1996) of the scientific Advisory Council to the Southern California
Earthquake Center. He has been a Distinguished Visiting Scientist at the Jet Propulsion
Laboratory, Pasadena, CA (1995-present), is currently an External Professor at the Santa Fe
Institute, and is a Fellow of both the American Physical Society (2005) and the American
Geophysical Union (2008). Recently, he was a co-winner of the NASA Software of the Year
Award (2012). John received his B.S.E from Princeton University (Magna Cum Laude, Phi
Beta Kappa, Tau Beta Pi), and M.S. (1973) and Ph.D. (1976) from the University of California
at Los Angeles. In addition to natural hazards and earthquakes, he also has professional
interests in forecasting, validation of forecasts, and quantitative finance. He currently coorganizes (along with Michael Maouboussin, Will Tracy and Martin Lebowitz) a yearly
meeting on risk for the Santa Fe Institute, often held at Morgan Stanley, Inc., in New
York. He teaches courses in Risk and Natural Disasters; Complex Systems; and Econophysics
and Quantitative Finance at the University of California, Davis.
About Me
The Open Hazards Group
Technical Team
•
John Rundle. Board Chair, Earthquake Forecasting. He is a frequent commentator on earthquake science on the
major national news networks CNN, FOX, and MSNBC.
•
Bill Graves. President, CEO, Board Member. Bill Graves has worked as the Director of the Systems Science Lab
and the Acting Director of Test and Measurements, Schlumberger. He was the first President and CEO of Cisco
Systems (1987-1988). Bill Graves earned a BS in Physics from MIT in 1968, served in the US Army Reserve from
1969-1975, and then completed a PhD in Physics at UC Berkeley in 1977.
•
James R. Holliday. Treasurer, Board Member, Forecast Validation and Software Development. James Holliday
is a Research Scientist in Physics at UC Davis, and former President of Xerasys, Inc. He earned a BS in Physics at
the University of Kansas in 1998 and a PhD in Physics at UC Davis in 2007.
•
Boris Jeremic. Geotechnical Engineering. Boris Jeremic is a Professor of Civil Engineering at UC Davis and a
Faculty Scientist at Lawrence Berkeley National Laboratory. He earned a Diploma Engineering degree in Civil
Engineering from Belgrade University in 1989 and then worked as a consulting engineer in Yugoslavia, Iraq and
Switzerland for three years. He completed post-graduate work at the University of Colorado where he earned
an MS in 1994 and a PhD in 1997 in Civil Engineering.
•
Sashi Kunnath. Earthquake and Structural Engineering. Sashi Kunnath is a Professor of Civil and Environmental
Engineering at UC Davis and Editor for the ASCE Journal of Structural Engineering. He is a Fellow of the
American Concrete Institute and a former Chair of the ASCE Committee on Seismic Effects. Sashi Kunnath
earned a B. Eng. in Civil Engineering at Bangalore University in 1980, a M. Eng. in Structural Engineering at the
University of Bangkok in 1982 and a PhD in Structural Engineering and Earthquake Engineering at SUNY Buffalo
in 1989. In 2008 he was awarded the ASCE Raymond Reese Research Prize for his work in modal adaptive
pushover analysis of building structures. He is also a registered Professional Engineer.
•
Kevin Mayeda. Seismologist. Kevin Mayeda is a Senior Scientist at Weston Geophysical and a Research Faculty
member at the UC Berkeley Seismology Laboratory. He was a Senior Seismologist at Lawrence Livermore
National Laboratory from 1992 to 2006. Dr. Mayeda earned a BS in Geophysics at UC Berkeley in 1987 and a
PhD in Geophysics at the University of Southern California in 1991.
•
Steven Ward. Seismologist. Steven Ward is a Research Geophysicist at the Institute of Geophysics and
Planetary Physics, UC Santa Cruz. He earned a BS in Physics at Bucknell University (1974) and a PhD in
Geophysics at Princeton (1978). Steven specializes in the quantification and simulation of natural hazards.
www.openhazards.com